Virtual Lane Segmentation From Vehicle Trajectories in Complex Traffic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in managing complex traffic situations, such as temporary obstacles and unclear road markings, as they rely on detection of road markings and static/dynamic objects, making it difficult to determine virtual lanes effectively.
Innovation Solution
A method for determining virtual traffic lanes using a history of movement data from multiple vehicles, which reconstructs lanes by generating a cloud of points and calculating proximity measurements to connect or separate points based on Mahalanobis distance, allowing for efficient filtering of outliers and improved precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous vehicles rely on detection of road markings and static/dynamic objects to manage traffic situations, then the vehicle can operate with standard detection systems, but it becomes difficult to determine virtual lanes effectively in complex situations such as temporary obstacles and unclear road markings
Solution Approach 1:
The patent introduces an intermediary system that collects and processes movement data from multiple vehicles to generate virtual lane information. This intermediary layer (the virtual lane determination system) mediates between the raw detection data and the autonomous vehicle's navigation decisions, enabling reliable operation in complex situations where direct detection of road markings fails.
Solution Approach 2:
The virtual lane determination system serves multiple functions: it reconstructs road geometry from movement data, identifies valid travel paths, filters outliers, and provides navigation guidance. This multi-functional approach allows a single system to address various complex traffic situations including temporary obstacles, unclear markings, and traffic jam maneuvers.
2Quantity of substance
If the system uses complete trajectories from multiple vehicles to determine virtual lanes, then more data is available for accurate lane reconstruction, but segments where vehicles change lanes must be discarded entirely, reducing data quality
Solution Approach 1:
The patent divides complete vehicle trajectories into smaller segments and processes them independently. This segmentation allows the system to retain useful segments even when parts of a trajectory are invalid (e.g., during lane changes). The system can discard only the problematic segments while preserving the rest of the data, thereby maintaining both quantity and precision.
Solution Approach 2:
The system applies different quality criteria to different segments of trajectories based on local conditions. Instead of uniformly discarding entire trajectories, it evaluates each segment's validity locally, keeping segments that conform to virtual lane geometry and discarding only those with outliers or inconsistencies. This local quality assessment preserves more useful data.
3Device complexity
If the system processes complete vehicle trajectories without segmentation, then the processing is simpler, but outliers cannot be efficiently filtered and the accuracy of virtual paths deteriorates
Solution Approach 1:
By segmenting trajectories into smaller units, the system enables efficient outlier filtering at the segment level. This segmentation increases processing complexity slightly but dramatically improves accuracy by allowing precise identification and removal of outlier segments without discarding entire trajectories. The segmented approach makes the filtering process computationally manageable.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Method for determining virtual lanes (VL1, VL2) for traffic flow on a carriageway of a road, on the basis of a history of motion data of at least one first and one second vehicle travelling on said carriageway, the method comprises the steps of: obtaining a first past trajectory (T1) of the first vehicle on the basis of the history of the motion data of the first vehicle; obtaining a second past trajectory (T2) of the second vehicle on the basis of the history of the motion data of the second vehicle; calculating a plurality of segments each comprising a set of points of the first and of the second trajectory, the points of said set together exhibiting a proximity measurement of less than a predetermined threshold; generating a virtual traffic flow lane on the carriageway on the basis of a grouping of at least some of the segments of said plurality of segments.